Games help spot and discuss bias in text to image AI outputs

Playing Bias: Epistemic Games for Interpreting Text-to-Image Models

Computers and SocietyComputer Vision and Pattern RecognitionHuman-Computer Interaction

Summary

Text-to-image AI systems don't just create images from words; they also show common ideas about what things look like in ways that can spread stereotypes. The authors designed a set of playful activities, called epistemic games, that help people carefully examine and question these biases in AI-generated images. These games encourage users to look closely at images, try changing prompts to explore ideas, and mix prompts to reveal hidden biases. When tested with art students, the games helped students make clearer, evidence-based observations about bias in AI images. The authors suggest these games can be used in classrooms to help people learn to recognize and talk about bias in AI visuals.

What this means in practice

  • For ai ethics teams: Use these games to identify and discuss bias in text-to-image model outputs during model evaluation and auditing.
  • For design educators: Integrate the games into art and design curricula to teach students how AI-generated images can reflect and reinforce stereotypes.

Authors

Amalia Foka

Abstract

Text-to-image (T2I) systems do not merely render prompts; they circulate defaults for "what X looks like," shaping recognition through pose, camera, costume, props, and setting. While technical audits quantify disparities, they rarely cultivate the literacy needed to argue how images assemble stereotypes and confer authority. This paper translates a three-stage, humanities-informed framework into a set of epistemic games that make this cultural work visible and debatable: Interpret (art-historical close looking with AI outputs), Interrogate (counter-intent prompting as artistic exploration), and Expose (creative misuse through paired prompts with a constant motif and a BiasSlot rule, in which only the bias cue varies). In a pilot with undergraduate fine-arts students, the games shifted discussion from impressionistic reactions to concise, warranted hypotheses grounded in visual evidence, supported by lightweight classroom instruments that enabled reproducibility and structured collaboration. We argue that such game-based approaches foster critical literacy across educational contexts, equipping learners to recognize, discuss, and contest how T2I models encode bias.
View PDFOpen arXiv